混合模型與影像細胞單元元素競爭法:應用在三維度電腦斷層掃描影像

Abstract

隨著電腦運算的能力增加,在取得和處理三維度影像變得可行.本論文研究目標是以混合模型和細胞單元元素競爭為基礎來分割三維度影像. 首先,分水嶺轉換法延伸到三維影像並且產生初始細胞單元.利用有限混合模型來模擬每一個細胞單元的平均影像強度值.在將起始細胞單元分類成前景和背景兩群.在相同群組並且相鄰的細胞單元合併.以每一個細胞單元的影像強度值的分配為基準的無母數檢定法,像兩個樣本的Kolomogrov-Smirnov (K-S)檢定,用來檢定不同群的相鄰細胞單元的合併或是分離. 臨床的研究上在執行三維度X-Ray電腦斷層掃描影像.測試的結果新的方法和有經驗的醫師比較後.確定了我們提出新方法的可行性.

As the increasing power of computing power, acquisition and processing of 3D images processing become feasible. This study is aimed to segment 3D images based on mixture modeling and cell competition. First of all, the watershed transform is extended to 3D images and generate initial cells. Finite mixtures are used to model the distributions of the mean intensities in cells. Then, initial cells are clustered into two groups of foreground and background. Neighboring cells in the same group are then merged. Nonparamtric tests based on the distributions of intensities in cells, like the two-sample Kolomogorov-Smirnov (K-S) tests, are used to merge or split neighboring cells that belong to different groups. Empirical studies are performed with the 3D CT X-ray images. The test results are compared with the segmentation results of a medical expert. The comparisons confirm the feasibility of this new and automatic approach.